Advertisement

Do Cultural and Psychosocial Factors Contribute to Type 2 Diabetes Risk? A Look Into Vancouver's South Asian Community

      Abstract

      Objectives

      South Asian immigrants are generally healthy upon arrival, but precipitously develop diabetes after immigration. Whether cultural and psychosocial factors contribute to diabetes risk in this ethnic minority group remains unclear. Existing prediction models focus primarily on clinical and lifestyle factors. This study explored whether nontraditional risk factors are incrementally predictive beyond traditional risk factors in this South Asian community.

      Methods

      In this cross-sectional study, we recruited 425 South Asian adults attending Sikh and Hindu temples in Metro Vancouver between July 2013 and June 2014. We measured traditional risk factors, including glycated hemoglobin (A1C), apolipoprotein B, systolic and diastolic blood pressure (BP), waist circumference, weight, body mass index (BMI), dietary patterns and physical activity level. Self-report questionnaires assessed cultural and psychosocial factors, including acculturation, dinnertime (timing of the evening meal), religion and depressive symptoms. We constructed a penalized multivariable linear model with A1C level using the least absolute shrinkage and selection operator (LASSO) approach to overcome issues of overfitting and reduce prediction error of previous diabetes prediction models.

      Results

      The LASSO model selected 24 risk factors for the optimal model to predict glycemic control. Results revealed that higher degree of acculturation (p=0.007), later dinnertime (p=0.01) and greater depressive symptoms (p=0.038) are important factors in diabetes risk in addition to traditional risk factors (fruit/vegetable/fibre intake, BMI and systolic BP).

      Conclusions

      Nontraditional factors, such as cultural practices and emotional functioning, are also important predictors of diabetes risk and should be considered when culturally tailoring diabetes prevention programs.

      Résumé

      Objectifs

      Les immigrants sud-asiatiques sont généralement en bonne santé à leur arrivée, mais développent rapidement le diabète après leur immigration. On ignore si des facteurs culturels et psychosociaux contribuent au risque de diabète dans ce groupe ethnique minoritaire. Les modèles de prédiction actuels portent principalement sur les facteurs cliniques et liés au mode de vie. La présente étude a permis d'examiner si les facteurs de risque non traditionnels sont des prédicteurs incrémentaux par rapport aux facteurs de risque traditionnels dans cette communauté sud-asiatique.

      Méthodes

      Dans la présente étude transversale, nous avons recruté 425 adultes sud-asiatiques fréquentant les temples sikhs et hindous de la région métropolitaine de Vancouver entre juillet 2013 et juin 2014. Nous avons mesuré les facteurs de risque traditionnels, dont l'hémoglobine glyquée (A1c), l'apolipoprotéine B, la pression artérielle (PA) systolique et diastolique, le périmètre abdominal, le poids, l'indice de masse corporelle (IMC), les habitudes alimentaires et le niveau d'activité physique. Les questionnaires auto-administrés ont permis d’évaluer les facteurs culturels et psychosociaux, dont l'acculturation, l'heure du souper (repas du soir), la religion et les symptômes de dépression. Nous avons élaboré un modèle de régression linéaire multiple pénalisée de la concentration de l’A1c à l'aide de l'approche LASSO (least absolute shrinkage and selection operator) pour surmonter les problèmes de surajustement et réduire l'erreur de prédiction des modèles de prédiction antérieurs du diabète.

      Résultats

      Le modèle LASSO a permis de sélectionner 24 facteurs de risque du modèle optimal pour prédire la régulation de la glycémie. En plus des facteurs de risque traditionnels (consommation de fruits, de légumes et de fibres, IMC et PA systolique), les résultats ont révélé qu'un degré supérieur d'acculturation (p = 0,007), une heure de souper plus tardive (p = 0,01) et des symptômes de dépression plus importants (p = 0,038) sont des facteurs importants du risque de diabète.

      Conclusions

      Les facteurs non traditionnels comme les pratiques culturelles et le fonctionnement émotionnel sont également des prédicteurs importants du risque de diabète et devraient être considérés lors de l'adaptation des programmes de prévention du diabète à la réalité culturelle.

      Keywords

      Mots clés

      • Traditional risk factors for diabetes (fruit/vegetable/fibre intake, body mass index and systolic blood pressure) were identified in this sample of South Asian adults.
      • Higher degree of acculturation, later dinnertime and greater depressive symptoms emerged as cultural and psychosocial risk factors for diabetes in this group.
      • Cultural practices and emotional functioning are also important predictors of diabetes risk and should be considered in diabetes prevention programs.

      Introduction

      In Canada, South Asians immigrants are the largest and most rapidly growing visible minority group (
      Statistics Canada
      Immigration and ethnocultural diversity in Canada.
      ) and bear a greater burden of type 2 diabetes mellitus than other ethnic groups (
      • Nie J.
      • Ardern C.
      Association between obesity and cardiometabolic health risk in Asian-Canadian sub-groups.
      ,
      • Rana A.
      • de Souza R.
      • Kandasamy S.
      • Lear S.
      • Anand S.
      Cardiovascular risk among South Asians living in Canada: A systematic review and meta-analysis.
      ). In fact, according to the Canadian Community Health Survey, from 2001–2012, South Asian men experienced the highest increase in diabetes (6.7% to 15.2%) compared with their Caucasian, Chinese and black counterparts (
      • Chiu M.
      • Maclagan L.
      • Tu J.
      • Shah B.
      Temporal trends in cardiovascular disease risk factors among white, South Asian, Chinese and black groups in Ontario, Canada, 2001 to 2012: A population-based study.
      ). Not only does this ethnic group have a higher risk for developing diabetes, but they also tend to be diagnosed at younger ages (
      • Chiu M.
      • Austin P.
      • Manuel D.
      • Shah B.
      • Tu J.
      Deriving ethnic-specific BMI cutoff points for assessing diabetes risk.
      ) and lower body mass index (BMI) (
      • Ntuk U.
      • Gill J.
      • Mackay D.
      • Sattar N.
      • Pell J.
      Ethnic-specific obesity cutoffs for diabetes risk: Cross-sectional study of 490,288 UK biobank participants.
      ). Moreover, South Asians more rapidly progress from prediabetes to diabetes compared with Caucasians (
      • Anjana R.M.
      • Rani C.S.
      • Deepa M.
      • et al.
      Incidence of diabetes and prediabetes and predictors of progression among Asian Indians: 10-year follow-up of the Chennai Urban Rural Epidemiology Study (CURES).
      ). Considering these statistics, the early identification of South Asians at risk for diabetes should be a public health imperative.
      Identifying the predictors of developing diabetes and prediabetes in the South Asian immigrant population can permit early detection and aggressive preventive interventions. However, there are few comprehensive analyses examining factors for developing diabetes in this population, and analytical techniques do not account for multiple testing, raising concern for overinflated and spurious associations. To date, studies have reported traditional clinical characteristics and lifestyle behaviours as risk factors for diabetes in the South Asian community. Specifically, South Asians have higher insulin resistance (
      • Rana A.
      • de Souza R.
      • Kandasamy S.
      • Lear S.
      • Anand S.
      Cardiovascular risk among South Asians living in Canada: A systematic review and meta-analysis.
      ,
      • Lear S.
      • Kohli S.
      • Bondy G.
      • Tchernof A.
      • Sniderman A.
      Ethnic variation in fat and lean body mass and the association with insulin resistance.
      ), higher percentage of visceral fat distribution (
      • Rana A.
      • de Souza R.
      • Kandasamy S.
      • Lear S.
      • Anand S.
      Cardiovascular risk among South Asians living in Canada: A systematic review and meta-analysis.
      ), greater beta cell dysfunction (
      • Kanaya A.M.
      • Herrington D.
      • Vittinghoff E.
      • et al.
      Understanding the high prevalence of diabetes in US south Asians compared with four racial/ethnic groups: The MASALA and MESA studies.
      ) and a greater genetic load for type 2 diabetes (
      • Ntuk U.
      • Gill J.
      • Mackay D.
      • Sattar N.
      • Pell J.
      Ethnic-specific obesity cutoffs for diabetes risk: Cross-sectional study of 490,288 UK biobank participants.
      ). South Asian persons also tend to consume a diet that may increase risk of developing diabetes [i.e. a diet high in saturated fats, carbohydrates and transfatty acids and low in mono- and polyunsaturated fats, fruits, vegetables and fibre (
      • Holmboe-Ottesen G.
      • Wandel M.
      Changes in dietary habits after migration and consequences for health: A focus on South Asians in Europe.
      ,
      • Raj S.
      • Ganganna P.
      • Bowering J.
      Dietary habits of Asian Indians in relation to length of residence the United States.
      )]. Finally, compared with Caucasians, South Asians report lower levels of physical activity (
      • Williams E.
      • Stamatakis E.
      • Chandola T.
      • Hamer M.
      Assessment of physical activity levels in South Asians in the UK: Findings from the Health Survey for England.
      ,
      • Fischbacher C.
      • Hunt S.
      • Alexander L.
      How physically active are South Asians in the United Kingdom? A literature review.
      ) and cardiorespiratory fitness (
      • Ghouri N.
      • Purves D.
      • McConnachie A.
      • Wilson J.
      • Gill J.
      • Sattar N.
      Lower cardiorespiratory fitness contributes to increased insulin resistance and fasting glycaemia in middle-aged South Asian compared with European men living in the UK.
      ).
      Although traditional risk factors play a significant role in diabetes risk, few studies of the South Asian community have examined the potential role of culture and psychosocial functioning in the development of diabetes. In fact, only one study of 899 South Asian adults living in the United States included cultural and psychosocial measures (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ). Specifically, they found participants who reported lower physical activity levels, greater screen time, higher long-term psychological burden and less adherence to monthly or yearly fasting practices to have increased risk for diabetes (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ). Although the investigation by Shah et al (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ) assessed clinical factors (e.g. waist circumference, visceral fat, blood pressure, cholesterol), these variables were not entered in final prediction models. Rather, multivariate analyses were limited to associations between diabetes risk and nonbiological/clinical factors (e.g. lifestyle, cultural and psychosocial factors). Our study explored the incremental role of cultural and psychosocial risk factors beyond traditional risk factors in the development of diabetes among South Asian immigrants living in Canada.

      Methods

      Setting and identification

      For this cross-sectional study, we obtained ethics approval from the University of British Columbia and Fraser Health Clinical Research Ethics Boards. We recruited a convenience sample of South Asian adults attending Sikh and Hindu temples in the Metro Vancouver area between July 2013 and June 2014. Patients were included in the study if they met the following criteria: 1) self-identify as South Asian, 2) ≥21 years of age, 3) have no previous diagnosis of diabetes, 4) speak Punjabi and/or English, 4) live in the Metro Vancouver area and 5) noted to be at increased risk of diabetes (score ≥5 out of 11 points) based on the 7-item American Diabetes Association diabetes risk test. The American Diabetes Association diabetes risk test assesses age, sex, history of gestational diabetes, family history of diabetes, diagnosis of hypertension, self-reported physical activity and weight.

      Outcomes and measurements

      Primary outcome

      Diabetes risk (e.g. glycated hemoglobin [A1C]) was obtained via venipuncture by a certified phlebotomist and analyzed using the CAPILLARYS Hb A1c kit (Roche Diagnostics, Indianapolis, Indiana, United States) at a central laboratory. Participants were placed in 1 of 3 diabetes risk categories: normal (A1C <42 mmol/mol; 6%), prediabetes (A1C of 42 to 46 mmol/mol; 6% to 6.4%) and diabetes (A1C ≥48 mmol/mol; 6.5%).

      Cultural measures

      Acculturation, languages spoken (English, Punjabi, Hindi, Gujarati, Urdu or other) and religion (Sikh, Hindu or other) were assessed via self-report. Acculturation was defined as the number of years living in Canada, with longer stay corresponding to a higher degree of acculturation. Dinnertime was also assessed via self-report and included the following item: On average, what time do you eat dinner (or supper) every evening?

      Psychosocial factors

      Depressive symptoms were measured using the Personal Health Inventory-9 (PHQ-9) (
      • Kroenke K.
      • Spitzer R.
      The PHQ-9: A new depression diagnostic and severity measure.
      ). The PHQ-9 has an overall accuracy of 85%, sensitivity of 75% and a specificity of 0.90% for major depressive disorder with scores of 5, 10, 15 and 20 corresponding to mild, moderate, moderately severe and severe depression, respectively (
      • Spitzer R.
      • Kroenke K.
      • Williams J.B.
      The Patient Health Questionnaire Primary Care Study Group. Validation and utility of a self-report version of PRIME-MD: The PHQ Primary Care Study.
      ).

      Clinical measures

      Seated resting blood pressure was taken 2 times using an automated blood pressure monitor (Omron BP 785; Omron Healthcare Co Ltd, Kyoto, Japan) with participants instructed to rest 1 min before and between readings. The average of the 2 readings was recorded. Weight was assessed using a digital weighing scale (Seca 874; Hammer Steindamm 3-25; Hamburg, Denmark) and height using a stadiometer (Seca 213; Hammer Steindamm 3-25). BMI was calculated from weight and height values (kg/m2). Waist circumference was measured using a flexible Seca waist tape at the umbilicus. Apolipoprotein B was analyzed using the UniCel DxC 600/800 Systems (Beckman Coulter; Brea, California, United States) and SYNCHRON System SPO Calibrator (Beckman Coulter; prior to May 2014). After May 2014, apolipoprotein B was analyzed using COBAS INTEGRA systems (Lower Mainland Pathology & Laboratory Medicine, Surrey, British Columbia, Canada).

      Lifestyle behaviour measures

      Participants completed a self-report survey administered by a trained bilingual research assistant. The survey included items that measured lifestyle behaviours (physical activity, sedentary behaviour, dietary patterns and other lifestyle habits).
      Fruit, vegetable and fibre (FVF) intake was assessed using a culturally adapted version of FVF screener initially developed by Block et al (
      • Block G.
      • Gillespie C.
      • Rosenbaum E.
      • Jenson C.
      A rapid food screener to assess fat and fruit and vegetable intake.
      ). On a 5-point Likert scale, participants were asked to rate the frequency with which they consumed items within the 8 separate food categories over the past month. A total score was calculated by adding up the numerical values for each food category. Fat intake was assessed using the Block et al culturally adapted version (
      • Block G.
      • Gillespie C.
      • Rosenbaum E.
      • Jenson C.
      A rapid food screener to assess fat and fruit and vegetable intake.
      ). Participants were asked to rate the frequency with which they consumed items within 18 food categories on a 5-point Likert scale. A total score was calculated by adding up the numerical value for each food category. The Block et al screeners (
      • Block G.
      • Gillespie C.
      • Rosenbaum E.
      • Jenson C.
      A rapid food screener to assess fat and fruit and vegetable intake.
      ) have been shown to correlate well with a full-length questionnaire for both fruit and vegetable servings (r=0.71) and dietary intake of total and saturated fat (r>0.60).
      Physical activity was assessed using the International Physical Activity Questionnaire – Short Form (IPAQ-SF) (
      • Craig C.L.
      • Marshall A.L.
      • Sjostrom M.
      • et al.
      International physical activity questionnaire: 12-country reliability and validity.
      ). The IPAQ-SF consists of 9 items that measure moderate to vigorous activity, walking, sedentary behaviour and screen time (e.g. television, computer). In a 12-country study, the IPAQ-SF was found to have good test-retest reliability (Spearman ρ=0.8) and concurrent validity between short and long forms (pooled p=0.67; 95% confidence interval, 0.64 to 0.70) and moderate criterion validity when compared with accelerometers (median p=0.30; 95% confidence interval, 0.23 to 0.36) (
      • Craig C.L.
      • Marshall A.L.
      • Sjostrom M.
      • et al.
      International physical activity questionnaire: 12-country reliability and validity.
      ).
      Sociodemographic background items included age, sex, country of origin, marital status, education level, household income and employment status.

      Statistical analysis

      Descriptive analyses examined frequencies and measures of central tendency for sociodemographic, clinical, behavioural and psychosocial variables. Univariate analyses were performed to explore the relationships between diabetes status (normal, prediabetes and diabetes) and individual factors. Associations between diabetes status and factors were assessed using Fisher exact test for categorical variables and analysis of variance for continuous variables.
      To identify important factors of glycemic control, a regularized regression with an L1 norm penalty, least absolute shrinkage and selection operator (LASSO), was performed with A1C as the dependent variable (

      Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc Series B Stat Methodol 1996;58:267–88.

      ). Recent research into selective inference has provided a means to perform valid statistical inference in the presence of selection bias, an element of the LASSO method (
      • Tibshirani R.J.
      • Taylor J.
      • Lockhart R.
      • Tibshirani R.
      Exact post-selection inference for sequential regression procedures.
      ). The selectiveInference R package (The R Project for Statistical Computing, Vienna, Austria) was used to calculate p values for the variables selected by the LASSO model (

      R. The R project for statistical computing. https://www.R-project.org. Accessed April 16, 2018.

      ).
      Because of the novelty of this approach, variable importance was also quantified using a bootstrap resampling framework. In each bootstrap sample, the LASSO model was fit to the resampled data, and the presence (or absence) of each variable in the model was recorded. This was done for 5,000 random samples, and the proportion of times each variable was selected by the LASSO procedure represents its importance. Because of the factor coding used by the LASSO model, all categorical variables were coded using dummy variables, and effect estimates should be interpreted as the effect of belonging to the category indicated vs all other categories for that variable while adjusting for all other variables in the model.
      The LASSO approach, a technique commonly associated with machine learning, was chosen over stepwise model building approaches for 2 main reasons (

      Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc Series B Stat Methodol 1996;58:267–88.

      ). First, stepwise approaches are susceptible to overfitting data and increased type 1 error from extensive multiple testing of independent variables. Second, standard linear regression can also provide misleading results when multicollinearity is present; when correlated variables are included in a model, their significance is often underestimated because standard errors are inflated. The LASSO approach avoids this by not relying on significance for inclusion in the final model and by implicitly including the most important variable among a group of correlated variables, or by including entire groups of variables that contribute information irrespective of the significance of individual independent variables. It is because of this feature of the LASSO approach that we have provided 2 metrics, significance and importance, to robustly quantify the relative value of each independent variable in the final prediction model. Here, importance refers to the relative value of each variable in the prediction model, whereas significance has the usual interpretation using p values provided by the selective inference method (
      • Tibshirani R.J.
      • Taylor J.
      • Lockhart R.
      • Tibshirani R.
      Exact post-selection inference for sequential regression procedures.
      ).
      An alternative approach, using directed acyclic graphs, has been used in a similar study (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ). This approach is another alternative to the stepwise regression approaches, but relies on extensive and untestable assumptions of the associations between variables. Furthermore, it requires separate models for every variable of interest. The LASSO approach assumes no causal structure among the variables and allows for the identification of single or groups of variables that contribute most to the model. All statistical tests were performed using R version 3.4.2 (

      R. The R project for statistical computing. https://www.R-project.org. Accessed April 16, 2018.

      ).

      Results

      Of the 551 participants screened, 425 (77%) were eligible and consented to participate. Of the 425 participants, 46 (11%) had undiagnosed type 2 diabetes (A1C ≥48 mmol/mol; 6.5%), 215 (51%) had prediabetes (A1C of 42 to 46 mmol/mol; 6% to 6.4%) and 164 (39%) had an A1C within normal range.
      Table 1 presents the sociodemographic characteristics of study participants in total and by diabetes risk status. More men (58%) than women participated in the study, ages ranged from 26 to 92 years (mean, 65±10 years) and years living in Canada ranged from 1 to 82 years (mean, 24±15 years). Almost all study participants were born in India, and most were married, with a high school education or less. More than 50% of study participants reported household income of <$40,000, whereas less than one-third of participants reported having full-time or part-time employment.
      Table 1Baseline characteristics of study participants by diabetes risk status
      CharacteristicOverall (N=425)Normal (n=164)Prediabetes (n=215)Diabetes (n=46)p value
      Sociodemographic background
       AgeMean ± SD65.32±10.4564.55±1165.73±10.0566.17±10.370.471
       Sex (male)Number missing (%)4 (1)2 (1)2 (1)0 (0)0.179
      n (%)245 (58)103 (63)115 (53)27 (59)
       EducationNumber missing (%)7 (2)5 (3)2 (1)0 (0)0.488
      <High school202 (48)69 (42)109 (51)24 (52)
      High school119 (28)46 (28)60 (28)13 (28)
      >High school97 (23)44 (27)44 (20)9 (20)
       Employed (full or part time)Number missing (%)6 (1)4 (2)2 (1)0 (0)0.062
      n (%)113 (27)52 (32)47 (22)14 (30)
       IncomeNumber missing (%)66 (16)30 (18)30 (14)6 (13)0.73
      <$20,00096 (23)34 (21)50 (23)12 (26)
      $20,000–$49,999148 (35)60 (37)75 (35)13 (28)
      ≥$50,000115 (27)40 (24)60 (28)15 (33)
       MarriedNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.275
      n (%)362 (85)144 (88)180 (84)38 (83)
       Country of originNumber missing (%)0 (0)0 (0)0 (0)0 (0)0.166
      India397 (93)152 (93)205 (95)40 (87)
      Other South Asia14 (3)5 (3)5 (2)4 (9)
      Diaspora14 (3)7 (4)5 (2)2 (4)
       Years living in CanadaMean ± SD23.89±15.4122.62±16.0724.45±14.6725.69±16.320.38
       ReligionNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.034
      Sikh376 (88)136 (83)195 (91)45 (98)
      Hindu38 (9)20 (12)17 (8)1 (2)
      Other6 (1)5 (3)1 (0)0 (0)
      Language proficiency
       EnglishNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.524
      n (%)250 (59)100 (61)121 (56)29 (63)
       PunjabiNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.089
      n (%)397 (93)147 (90)205 (95)45 (98)
       HindiNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.898
      n (%)203 (48)80 (49)101 (47)22 (48)
       GujaratiNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.858
      n (%)15 (4)6 (4)7 (3)2 (4)
       UrduNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.421
      n (%)42 (10)18 (11)22 (10)2 (4)
       OtherNumber missing (%)5 (1)3 (2)2 (1)0 (0)0.393
      n (%)13 (3)7 (4)6 (3)0 (0)
      Clinical variables
       ApoB (g/L)Mean ± SD0.99±0.250.99±0.240.99±0.271.05±0.250.364
      Systolic BP (mmHg)Mean ± SD134.25±17.32132.43±16.64134.2±16.99140.91±19.810.013
      Diastolic BP (mmHg)Mean ± SD80.41±10.0180.04±9.9180.05±9.6483.41±11.620.098
       Weight (kg)Mean ± SD76.03±12.9174.44±11.3976.23±13.5780.73±13.880.013
       BMI (kg/m2)Mean ± SD28.14±3.9527.32±3.8228.42±3.9729.74±3.7<0.001
       Waist circumference (cm)Mean ± SD101.56±9.92100.25±9.39101.68±10.07105.69±10.130.004
      Lifestyle behaviours
       Total fruit vegetable fibre scoreMean ± SD14.77±5.0315.43±5.3414.42±4.7514±4.990.084
       Total fat scoreMean ± SD13.39±5.7612.89±5.913.76±5.5413.46±6.230.344
       Physical activity—moderate or vigorousNumber missing (%)0.255
      Inactive140 (33)58 (35)64 (30)18 (39)
      Minimally active218 (51)75 (46)121 (56)22 (48)
      HEPA active67 (16)31 (19)30 (14)6 (13)
       MET minutesMean ± SD1,336.05±1,282.221,416.53±1,440.251,288.08±1,126.381,273.34±1,380.30.591
       Sedentary time (min)Mean ± SD274.03±204.74252.37±179.4284±210.28304.67±254.890.185
       Television or screen time (min)Mean ± SD119.27±99.79105.33±92.86123.36±97149±126.970.024
       Dinner timeMean ± SD07:12 pm (1.19 h)07:14 pm (1.29 h)07:09 pm (1.12 h)07:20 pm (1.1 h)0.623
       SmokingNumber missing (%)6 (1)4 (2)2 (1)0 (0)0.136
      n (%)9 (2)5 (3)2 (1)2 (4)
      Psychological variables
       PHQ-9 scoreMean ± SD2.35±3.772.19±4.032.36±3.322.83±4.720.608
      ApoB, apolipoprotein B; BMI, body mass index; BP, blood pressure; HEPA, health-enhancing physical activity; MET, metabolic equivalent of task; PHQ-9, Personal Health Inventory-9.
      In univariate analyses, increased BMI, weight, waist circumference and systolic blood pressure were significantly associated with having diabetes. Two nonclinical variables (affiliation with Sikh religion vs other religions and increased total television and/or computer screen time) emerged as significantly associated with diabetes. There were no other significant differences between those with diabetes, prediabetes and normal glycemic status.
      Multivariate analysis was performed using the LASSO technique. Figure 1 provides a graphical presentation of the importance of each variable in the LASSO model as assessed through the bootstrap procedure. The colouring in this figure illustrates whether the risk factor was selected by the LASSO procedure. Figure 2 illustrates the coefficient paths of the variables selected in the final model for different levels of penalization (lambda). The most important risk factors diverge from the 0 (horizontal) line for the highest penalization; the relative impact of each risk factor on the prediction of A1C is represented by the distance from the 0 (horizontal) line as the penalty approaches zero (moves left). Penalty (left side of the plot) is equivalent to an unpenalized, standard, multivariable regression model. The vertical dashed line represents the optimal penalization as selected by a leave-one-out cross-validation procedure. This level of penalization was used for all presented model results.
      Figure thumbnail gr1
      Figure 1Bootstrap assessed variable importance coloured by whether each variable was selected in the full LASSO model. ApoB, apolipoprotein B; BMI, body mass index; FVF, fruit, vegetable and fibre; HEPA, health-enhancing physical activity; IPAQ, International Physical Activity Questionnaire – Short Form; L, linear; LASSO, least absolute shrinkage and selection operator; MET, metabolic equivalent of task; PHQ, Personal Health Inventory-9; Q, quadratic.
      Figure thumbnail gr2
      Figure 2Plot of coefficient paths in the least absolute shrinkage and selection operator model. Variables in the legend are presented in the same order they enter the model and are coloured by their coefficient sign at the optimal penalization selection by the model. BMI, body mass index; FVF, fruit, vegetable and fibre; HEPA, health-enhancing physical activity; IPAQ, International Physical Activity Questionnaire – Short Form; MET, metabolic equivalent of task; PHQ, Personal Health Inventory-9; Q, quadratic.
      In assessing the relative value of each predictor in the model, the combination of a high bootstrap variable importance score (≥90%) and p<0.05 indicate variables that are strong risk factors, while accounting for other independent variables.
      Thirty-three variables were included as potential risk factors in the LASSO model. When coded using dummy variables, the total number of potential variables in the model increased to 41. It is also important to note that, because the survey included each language as a checkbox, all language variables were coded as binary (true or false) variables.
      Table 2 provides a tabular presentation of the results from this analysis. In contrast with the univariate analysis, total FVF was selected in every sample (variable importance 100%) and was an important predictor of A1C (p=0.001). Higher total FVF intake was associated with lower A1C when adjusting for all other variables.
      Table 2Variables selected by least absolute shrinkage and selection operator model, coefficient estimates, variable importance score and p values
      VariableEstimateVariable importance (%)p value
      Total FVF−0.0951000.001
      BMI0.081920.003
      Years in Canada0.061930.007
      Dinnertime
      Variables whose importance is unclear.
      0.055890.010
      Speaks other language
      Variables whose importance is unclear.
      −0.038850.037
      PHQ score
      Variables whose importance is unclear.
      0.044830.038
      Systolic blood pressure
      Variables whose importance is unclear.
      0.055870.041
      Size of household
      Variables whose importance is unclear.
      0.042880.048
      Gujarati speaking0.042850.063
      Age0.043910.084
      Country of origin—Diaspora0.027720.116
      Religion—other−0.032840.147
      IPAQ category—HEPA active0.015600.197
      Total fat−0.023770.204
      Religion—Sikh0.022750.213
      Television hours0.019730.271
      Education level—Q0.011660.287
      Punjabi speaking0.03790.310
      IPAQ category—inactive−0.0041450.338
      Sex (male)−0.012590.358
      Hindi speaking0.00056550.426
      Income—Q0.0049600.511
      Total MET0.015560.784
      Country of origin—India−0.0063460.809
      BMI, body mass index; FVF, fruit, vegetable and fibre; HEPA, health-enhancing physical activity; IPAQ, International Physical Activity Questionnaire – Short Form; MET, metabolic equivalent of task; PHQ, Personal Health Inventory-9; Q, quadratic.
      Variables whose importance is unclear.
      As expected, BMI was selected in the LASSO model and was important by both metrics of importance (variable importance ≥90% and p≤0.05). Years in Canada also had a variable importance score ≥90% and p value <0.05. Age had a variable importance score ≥90%, but its associated p value did not meet the 0.05 cutoff (p=0.084). Dinnertime, speaks other language, PHQ-9 score, systolic blood pressure and size of household had significant p values but did not have variable importance scores ≥90%. The disagreement between these variables' bootstrap importance scores and p values indicates that they will create superior predictions of A1C, but do not contribute as much information as the variables that had both an importance score >90% and p value <0.05. The remaining variables selected by the LASSO model did not meet either the variable importance score cutoff of ≥90% or the p value cutoff of <0.05.

      Discussion

      To our knowledge, this study is the first to screen for and identify cultural and psychosocial predictors of diabetes risk among South Asian immigrants living in Canada. Most notably, traditional and nontraditional (cultural and psychosocial) factors were considered collectively in our prediction model. Based on the A1C values in this cohort, 11% had undiagnosed diabetes, whereas 51% had prediabetes. These proportions differ from the American study of South Asian adults, which reported 25% and 33% of the sample to have undiagnosed diabetes and prediabetes, respectively (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ). Differences, notwithstanding, given that South Asians progress from prediabetes to diabetes at a faster rate (up to 18% per year) (
      • Ramachandran A.
      • Snehalatha C.
      • Mary S.
      • Mukesh B.
      • Bhaskar A.
      • Vijay V.
      The Indian Diabetes Prevention Programme shows that lifestyle modification and metformin prevent type 2 diabetes in Asian Indian subjects with impaired glucose tolerance (IDPP-1).
      ) than other ethnic groups (up to 11% per year) (
      • Pan X.R.
      • Li G.W.
      • Hu Y.H.
      • et al.
      Effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance: The Da Qing IGT and Diabetes Study.
      ,
      • Knowler W.C.
      • Barrett-Connor E.
      • Fowler S.E.
      • et al.
      Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin.
      ,
      • Tuomilehto J.
      • Lindström J.
      • Eriksson J.G.
      • et al.
      Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance.
      ), community-wide screening, particularly targeting those at higher risk, can be instrumental for early detection and aggressive prevention efforts (
      • Sattar N.
      • Gill J.
      Type 2 diabetes in migrant south Asians: Mechanisms, mitigation, and management.
      ).
      Not surprisingly, BMI, a well-known and traditional predictor of diabetes development in the South Asian community, was one of the most important risk factors identified in this study. Recently, new cutoff values for BMI were established for this ethnic group. Although BMIs ≥25 and ≥30 kg/m2 were traditionally classified as increased risk and high risk for diabetes, respectively, the United Kingdom's National Institute for Health and Care Excellence recommends that these cutoff values be lowered to ≥23 kg/m2 (increased risk) and ≥27.5 kg/m2 (high risk) (
      National Institute for Health and Care Excellence
      BMI: Preventing ill health and premature death in black, Asian and other minority ethnic groups.
      ) for South Asians. Given that our study reported a mean BMI of 27.3 kg/m2 for the normal group and ≥28.4 kg/m2 for the prediabetes and diabetes groups, our sample, as a whole, would be categorized as high risk. These new ethnic-specific cutoff values and risk stratification guidelines for BMI may strategically function as an early warning signal for South Asians to take action preemptively.
      Acculturation and FVF intake also emerged as important risk factors for diabetes in our study. Although previous research has found that migration and urbanization can exacerbate dietary patterns after South Asians immigrate to Western countries (
      • Raj S.
      • Ganganna P.
      • Bowering J.
      Dietary habits of Asian Indians in relation to length of residence the United States.
      ), Lesser et al (
      • Lesser I.
      • Gasevic D.
      • Lear S.
      The association between acculturation and dietary patterns of South Asian immigrants.
      ) reported fruit and vegetable consumption to increase as a function of years living in Canada. Considering that we found diabetes risk to be positively correlated with acculturation, but negatively correlated with FVF intake, the relationship among these 3 variables is clearly complex and warrants further investigation.
      Another novel cultural association that emerged as a risk factor for diabetes was delayed dinnertime (i.e. timing of the evening meal). According to Simmons and Williams (
      • Simmons D.
      • Williams R.
      Dietary practices among Europeans and different South Asian groups in Coventry.
      ), South Asians typically consume their evening meal 2 to 3 h later than their Caucasian counterparts. Recently, a proof of concept investigation in non–South Asian participants found that delayed eating was associated with weight gain and elevated insulin levels (
      • Goel N.
      • Hopkins C.
      • Ruggieri M.
      • Ahima R.
      • Allison K.
      Delayed eating adversely impacts weight and metabolism compared with daytime eating in normal weight adults.
      ). This association should be validated as a putative marker for accelerated dysglycemia in South Asian populations.
      Although our multivariate model identified depressive symptomology as a predictor of diabetes risk at p<0.038, bootstrap analysis only identified this variable as important 83% of the time. As such, the role of depression in diabetes risk, as it pertains to this study, is not definitive. However, given that the study by Shah et al (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ) shows a positive correlation between psychological burden and diabetes prevalence, these types of psychosocial risk factors deserve greater attention. It is possible that the PHQ-9, a measure designed for diagnosing depression rather than assessing a wider range of depressive symptoms, was not sensitive enough to capture a relationship between depression and diabetes risk, if it did exist. An alternative measure, such as the Hamilton Depression Rating Scale, which assesses for somatic symptoms, may be more appropriate for some ethnic populations.
      Considering the wide heterogeneity in the South Asian population, findings from this prediction model have important implications for culturally tailoring prevention initiatives targeting this specific community. At least 3 large-scale diabetes prevention studies have been conducted with the South Asian population. The Indian Diabetes Prevention Program recruited middle-class men and women working in service organizations and associated families (
      • Ramachandran A.
      • Snehalatha C.
      • Mary S.
      • Mukesh B.
      • Bhaskar A.
      • Vijay V.
      The Indian Diabetes Prevention Programme shows that lifestyle modification and metformin prevent type 2 diabetes in Asian Indian subjects with impaired glucose tolerance (IDPP-1).
      ). The Prevention of Diabetes and Obesity in South Asians (PODOSA) trial (
      • Bhopal R.S.
      • Douglas A.
      • Wallia S.
      • et al.
      Effect of a lifestyle intervention on weight change in south Asian individuals in the UK at high risk of type 2 diabetes: A family-cluster randomised controlled trial.
      ) recruited men and women through the National Health Service, community organizations and word of mouth in the United Kingdom, and a mobile phone messaging trial recruited men employed in public and private sector industrial units in Southeast India (
      • Ramachandran A.
      • Snehalatha C.
      • Ram J.
      • et al.
      Effectiveness of mobile phone messaging in prevention of type 2 diabetes by lifestyle modification in men in India: A prospective, parallel-group, randomised controlled trial.
      ). These trials found lifestyle interventions to reduce the risk of developing diabetes by 28.4% to 34%. The magnitude of improvements, however, was more modest than studies conducted in the United States, Finland and China (
      • Pan X.R.
      • Li G.W.
      • Hu Y.H.
      • et al.
      Effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance: The Da Qing IGT and Diabetes Study.
      ,
      • Knowler W.C.
      • Barrett-Connor E.
      • Fowler S.E.
      • et al.
      Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin.
      ,
      • Tuomilehto J.
      • Lindström J.
      • Eriksson J.G.
      • et al.
      Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance.
      ). Perhaps to augment treatment effects in the South Asian population, we need to further personalize interventions at the neighborhood and community level. In other words, to increase cultural relevance and health impact, these findings need to be integrated in the intervention tailoring process. For instance, a future prevention program for this faith-based, South Asian community should place a greater emphasis on discussing the effect of acculturation on lifestyle behaviours, increasing fruit and vegetable intake, moving dinnertime earlier in the evening, developing effective coping strategies for psychosocial stress and educating about depression along with various treatment options.
      There are some limitations to this study. Given that participants were recruited predominantly from Sikh temples, findings may not be generalizable to the larger South Asian population in Canada. South Asians represent a heterogeneous population regarding education level, income, employment and literacy. In fact, 76% of our sample had a high school degree or lower, whereas >80% of the cohort that Shah et al (
      • Shah A.
      • Vittinghoff E.
      • Kandula N.
      • Srivastava S.
      • Kanaya A.
      Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
      ) recruited had a bachelor's degree or higher. In addition, we measured dietary patterns using brief food screeners rather than more rigorous methods, such as food frequency questionnaires and 24-h recall. Therefore, results need to be interpreted with caution. Although the strength of our study was the inclusion of risk factors across multiple domains, there still remained variables we did not measure, such as insulin resistance. Although expensive, future studies should include all putative biological, lifestyle, environmental, psychosocial and cultural risk factors to optimize prediction models. Finally, lifestyle behaviours (e.g. physical activity, diet) were assessed using self-report surveys that are subject to recall and social desirability bias.

      Conclusions

      Considering the rapidly growing rate of type 2 diabetes in Canada's South Asian population, public health efforts may have a greater impact on interventions focused on early identification and prevention. Risk models may need to also include cultural and psychosocial factors for early identification of diabetes risk in South Asian immigrant populations. Prevention programs that target the unique set of modifiable risk factors most salient in the target community will likely produce the largest health-related benefits. Based on the increased vulnerability to diabetes in the South Asian population, increased attention is needed regarding routine community-wide screening, early identification of high-risk individuals and timely initiation of lifestyle change interventions.

      Acknowledgments

      We thank the very dedicated research staff who assisted with clinical assessments. This research was supported by a grant sponsored by the Vancouver Foundation , Heart and Stroke Foundation , Vancouver General Hospital-University of British Columbia Hospital Foundation and the Azad and Yasmin Shamji Family. This research was also supported, in part, by an unrestricted educational grant from Sanofi-Aventis . The funding sources had no role in the study design; data collection; administration of the interventions; analysis, interpretation or reporting of data or decision to submit the findings for publication.

      Author Disclosures

      Conflicts of interest: None.

      Author Contributions

      The principal investigators T.S.T. and P.B. had full access to all the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis. T.S.T. contributed to study conception and design, study implementation, data analysis and interpretation and manuscript preparation. K.H. contributed to data analysis and interpretation and manuscript preparation. P.S. contributed to study implementation. P.B. contributed to study conception and design, study implementation and manuscript preparation. N.K. contributed to study conception and design, data interpretation and manuscript preparation.

      References

        • Statistics Canada
        Immigration and ethnocultural diversity in Canada.
        • Nie J.
        • Ardern C.
        Association between obesity and cardiometabolic health risk in Asian-Canadian sub-groups.
        PLoS One. 2014; 9: e107548
        • Rana A.
        • de Souza R.
        • Kandasamy S.
        • Lear S.
        • Anand S.
        Cardiovascular risk among South Asians living in Canada: A systematic review and meta-analysis.
        CMAJ Open. 2014; 2: E183-E191
        • Chiu M.
        • Maclagan L.
        • Tu J.
        • Shah B.
        Temporal trends in cardiovascular disease risk factors among white, South Asian, Chinese and black groups in Ontario, Canada, 2001 to 2012: A population-based study.
        BMJ Open. 2015; 5: e007232
        • Chiu M.
        • Austin P.
        • Manuel D.
        • Shah B.
        • Tu J.
        Deriving ethnic-specific BMI cutoff points for assessing diabetes risk.
        Diabetes Care. 2011; 34: 1741-1748
        • Ntuk U.
        • Gill J.
        • Mackay D.
        • Sattar N.
        • Pell J.
        Ethnic-specific obesity cutoffs for diabetes risk: Cross-sectional study of 490,288 UK biobank participants.
        Diabetes Care. 2014; 37: 2500-2507
        • Anjana R.M.
        • Rani C.S.
        • Deepa M.
        • et al.
        Incidence of diabetes and prediabetes and predictors of progression among Asian Indians: 10-year follow-up of the Chennai Urban Rural Epidemiology Study (CURES).
        Diabetes Care. 2015; 38: 1441-1448
        • Lear S.
        • Kohli S.
        • Bondy G.
        • Tchernof A.
        • Sniderman A.
        Ethnic variation in fat and lean body mass and the association with insulin resistance.
        J Clin Endocrinol Metab. 2009; 94: 4696-4702
        • Kanaya A.M.
        • Herrington D.
        • Vittinghoff E.
        • et al.
        Understanding the high prevalence of diabetes in US south Asians compared with four racial/ethnic groups: The MASALA and MESA studies.
        Diabetes Care. 2014; 37: 1621-1628
        • Holmboe-Ottesen G.
        • Wandel M.
        Changes in dietary habits after migration and consequences for health: A focus on South Asians in Europe.
        Food Nutr Res. 2012; 56: 18891
        • Raj S.
        • Ganganna P.
        • Bowering J.
        Dietary habits of Asian Indians in relation to length of residence the United States.
        J Am Diet Assoc. 1999; 99: 1106-1108
        • Williams E.
        • Stamatakis E.
        • Chandola T.
        • Hamer M.
        Assessment of physical activity levels in South Asians in the UK: Findings from the Health Survey for England.
        J Epidemiol Community Health. 2010; 65: 517-521
        • Fischbacher C.
        • Hunt S.
        • Alexander L.
        How physically active are South Asians in the United Kingdom? A literature review.
        J Public Health (Oxf). 2004; 26: 250-258
        • Ghouri N.
        • Purves D.
        • McConnachie A.
        • Wilson J.
        • Gill J.
        • Sattar N.
        Lower cardiorespiratory fitness contributes to increased insulin resistance and fasting glycaemia in middle-aged South Asian compared with European men living in the UK.
        Diabetologia. 2013; 56: 2238-2249
        • Shah A.
        • Vittinghoff E.
        • Kandula N.
        • Srivastava S.
        • Kanaya A.
        Correlates of prediabetes and type II diabetes in US South Asians: Findings from the Mediators of Atherosclerosis in South Asians Living in America (MASALA) study.
        Ann Epidemiol. 2015; 25: 77-83
        • Kroenke K.
        • Spitzer R.
        The PHQ-9: A new depression diagnostic and severity measure.
        Psychiatr Ann. 2002; 32: 509-515
        • Spitzer R.
        • Kroenke K.
        • Williams J.B.
        The Patient Health Questionnaire Primary Care Study Group. Validation and utility of a self-report version of PRIME-MD: The PHQ Primary Care Study.
        JAMA. 1999; 282: 1737-1744
        • Block G.
        • Gillespie C.
        • Rosenbaum E.
        • Jenson C.
        A rapid food screener to assess fat and fruit and vegetable intake.
        Am J Prev Med. 2000; 18: 284-288
        • Craig C.L.
        • Marshall A.L.
        • Sjostrom M.
        • et al.
        International physical activity questionnaire: 12-country reliability and validity.
        Med Sci Sports Exerc. 2003; 25: 1381-1395
      1. Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc Series B Stat Methodol 1996;58:267–88.

        • Tibshirani R.J.
        • Taylor J.
        • Lockhart R.
        • Tibshirani R.
        Exact post-selection inference for sequential regression procedures.
        J Am Stat Assoc. 2016; 111: 600-620
      2. R. The R project for statistical computing. https://www.R-project.org. Accessed April 16, 2018.

        • Ramachandran A.
        • Snehalatha C.
        • Mary S.
        • Mukesh B.
        • Bhaskar A.
        • Vijay V.
        The Indian Diabetes Prevention Programme shows that lifestyle modification and metformin prevent type 2 diabetes in Asian Indian subjects with impaired glucose tolerance (IDPP-1).
        Diabetologia. 2006; 49: 289-297
        • Pan X.R.
        • Li G.W.
        • Hu Y.H.
        • et al.
        Effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance: The Da Qing IGT and Diabetes Study.
        Diabetes Care. 1997; 20: 537-544
        • Knowler W.C.
        • Barrett-Connor E.
        • Fowler S.E.
        • et al.
        Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin.
        N Engl J Med. 2002; 346: 393-403
        • Tuomilehto J.
        • Lindström J.
        • Eriksson J.G.
        • et al.
        Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance.
        N Engl J Med. 2001; 344: 1343-1350
        • Sattar N.
        • Gill J.
        Type 2 diabetes in migrant south Asians: Mechanisms, mitigation, and management.
        Lancet Diabetes Endocrinol. 2015; 3: 1004-1016
        • National Institute for Health and Care Excellence
        BMI: Preventing ill health and premature death in black, Asian and other minority ethnic groups.
        http://www.nice.org.uk/guidance/ph46
        Date accessed: April 16, 2018
        • Lesser I.
        • Gasevic D.
        • Lear S.
        The association between acculturation and dietary patterns of South Asian immigrants.
        PLoS One. 2014; 9: e88495
        • Simmons D.
        • Williams R.
        Dietary practices among Europeans and different South Asian groups in Coventry.
        Br J Nutr. 1997; 78: 5
        • Goel N.
        • Hopkins C.
        • Ruggieri M.
        • Ahima R.
        • Allison K.
        Delayed eating adversely impacts weight and metabolism compared with daytime eating in normal weight adults.
        Sleep. 2017; 40: A24-A25
        • Bhopal R.S.
        • Douglas A.
        • Wallia S.
        • et al.
        Effect of a lifestyle intervention on weight change in south Asian individuals in the UK at high risk of type 2 diabetes: A family-cluster randomised controlled trial.
        Lancet Diabetes Endocrinol. 2014; 2: 218-227
        • Ramachandran A.
        • Snehalatha C.
        • Ram J.
        • et al.
        Effectiveness of mobile phone messaging in prevention of type 2 diabetes by lifestyle modification in men in India: A prospective, parallel-group, randomised controlled trial.
        Lancet Diabetes Endocrinol. 2013; 1: 191-198